# Rethinking Learning-Based Influence Maximization: Simple Neural Surrogates and Native Discrete Search

> Source: <https://www.machinebrief.com/news/rethinking-learning-based-influence-maximization-simple-neur-ygm9>
> Published: 2026-08-11 04:00:00+00:00

arXiv:2608.08406v1 Announce Type: new
Abstract: Existing learning-based influence maximization frameworks rely heavily on complex neural architectures and continuous optimization over seed representations. We challenge this paradigm with SIMBA, a diffusion-model-agnostic framework pairing a lightweight neural surrogate with direct discrete search. SIMBA introduces three key components: 1) uniformly anchored node embeddings that eliminate initialization noise and encourage learning driven by graph topology and diffusion pattern, 2) a shallow two-layer graph neural network surrogate predicting final infection states, and 3) batched multi-swap simulated annealing that explores combinatorial seed space without gradients or continuous relaxation. By shifting compute from complex representation learning to effective discrete search, SIMBA drastically cuts time-to-solution while achieving superior influence spread and data efficiency. Our code is available at https://github.com/yl489/rethink-IM.